io_uring: move struct io_kiocb from task_struct to io_uring_task

Rather than store the task_struct itself in struct io_kiocb, store
the io_uring specific task_struct. The life times are the same in terms
of io_uring, and this avoids doing some dereferences through the
task_struct. For the hot path of putting local task references, we can
deref req->tctx instead, which we'll need anyway in that function
regardless of whether it's local or remote references.

This is mostly straight forward, except the original task PF_EXITING
check needs a bit of tweaking. task_work is _always_ run from the
originating task, except in the fallback case, where it's run from a
kernel thread. Replace the potentially racy (in case of fallback work)
checks for req->task->flags with current->flags. It's either the still
the original task, in which case PF_EXITING will be sane, or it has
PF_KTHREAD set, in which case it's fallback work. Both cases should
prevent moving forward with the given request.

Signed-off-by: Jens Axboe <[email protected]>

Author: axboe

include/linux/io_uring/cmd.h

@@ -110,7 +110,7 @@ static inline void io_uring_cmd_complete_in_task(struct io_uring_cmd *ioucmd,
 
 static inline struct task_struct *io_uring_cmd_get_task(struct io_uring_cmd *cmd)
 {
-	return cmd_to_io_kiocb(cmd)->task;
+	return cmd_to_io_kiocb(cmd)->tctx->task;
 }
 
 #endif /* _LINUX_IO_URING_CMD_H */

include/linux/io_uring_types.h

@@ -84,6 +84,7 @@ struct io_uring_task {
 	/* submission side */
 	int				cached_refs;
 	const struct io_ring_ctx 	*last;
+	struct task_struct		*task;
 	struct io_wq			*io_wq;
 	struct file			*registered_rings[IO_RINGFD_REG_MAX];
 
@@ -625,7 +626,7 @@ struct io_kiocb {
 	struct io_cqe			cqe;
 
 	struct io_ring_ctx		*ctx;
-	struct task_struct		*task;
+	struct io_uring_task		*tctx;
 
 	union {
 		/* stores selected buf, valid IFF REQ_F_BUFFER_SELECTED is set */

io_uring/cancel.c

@@ -205,7 +205,7 @@ int io_async_cancel(struct io_kiocb *req, unsigned int issue_flags)
 		.opcode	= cancel->opcode,
 		.seq	= atomic_inc_return(&req->ctx->cancel_seq),
 	};
-	struct io_uring_task *tctx = req->task->io_uring;
+	struct io_uring_task *tctx = req->tctx;
 	int ret;
 
 	if (cd.flags & IORING_ASYNC_CANCEL_FD) {

io_uring/fdinfo.c

@@ -203,7 +203,7 @@ __cold void io_uring_show_fdinfo(struct seq_file *m, struct file *file)
 
 		hlist_for_each_entry(req, &hb->list, hash_node)
 			seq_printf(m, "  op=%d, task_works=%d\n", req->opcode,
-					task_work_pending(req->task));
+					task_work_pending(req->tctx->task));
 	}
 
 	if (has_lock)

io_uring/io_uring.c

@@ -206,7 +206,7 @@ bool io_match_task_safe(struct io_kiocb *head, struct io_uring_task *tctx,
 {
 	bool matched;
 
-	if (tctx && head->task->io_uring != tctx)
+	if (tctx && head->tctx != tctx)
 		return false;
 	if (cancel_all)
 		return true;
@@ -407,11 +407,8 @@ static void io_clean_op(struct io_kiocb *req)
 		kfree(req->apoll);
 		req->apoll = NULL;
 	}
-	if (req->flags & REQ_F_INFLIGHT) {
-		struct io_uring_task *tctx = req->task->io_uring;
-
-		atomic_dec(&tctx->inflight_tracked);
-	}
+	if (req->flags & REQ_F_INFLIGHT)
+		atomic_dec(&req->tctx->inflight_tracked);
 	if (req->flags & REQ_F_CREDS)
 		put_cred(req->creds);
 	if (req->flags & REQ_F_ASYNC_DATA) {
@@ -425,7 +422,7 @@ static inline void io_req_track_inflight(struct io_kiocb *req)
 {
 	if (!(req->flags & REQ_F_INFLIGHT)) {
 		req->flags |= REQ_F_INFLIGHT;
-		atomic_inc(&req->task->io_uring->inflight_tracked);
+		atomic_inc(&req->tctx->inflight_tracked);
 	}
 }
 
@@ -514,7 +511,7 @@ static void io_prep_async_link(struct io_kiocb *req)
 static void io_queue_iowq(struct io_kiocb *req)
 {
 	struct io_kiocb *link = io_prep_linked_timeout(req);
-	struct io_uring_task *tctx = req->task->io_uring;
+	struct io_uring_task *tctx = req->tctx;
 
 	BUG_ON(!tctx);
 	BUG_ON(!tctx->io_wq);
@@ -529,7 +526,7 @@ static void io_queue_iowq(struct io_kiocb *req)
 	 * procedure rather than attempt to run this request (or create a new
 	 * worker for it).
 	 */
-	if (WARN_ON_ONCE(!same_thread_group(req->task, current)))
+	if (WARN_ON_ONCE(!same_thread_group(tctx->task, current)))
 		atomic_or(IO_WQ_WORK_CANCEL, &req->work.flags);
 
 	trace_io_uring_queue_async_work(req, io_wq_is_hashed(&req->work));
@@ -678,17 +675,17 @@ static void io_cqring_do_overflow_flush(struct io_ring_ctx *ctx)
 }
 
 /* must to be called somewhat shortly after putting a request */
-static inline void io_put_task(struct task_struct *task)
+static inline void io_put_task(struct io_kiocb *req)
 {
-	struct io_uring_task *tctx = task->io_uring;
+	struct io_uring_task *tctx = req->tctx;
 
-	if (likely(task == current)) {
+	if (likely(tctx->task == current)) {
 		tctx->cached_refs++;
 	} else {
 		percpu_counter_sub(&tctx->inflight, 1);
 		if (unlikely(atomic_read(&tctx->in_cancel)))
 			wake_up(&tctx->wait);
-		put_task_struct(task);
+		put_task_struct(tctx->task);
 	}
 }
 
@@ -1207,7 +1204,7 @@ static inline void io_req_local_work_add(struct io_kiocb *req,
 
 static void io_req_normal_work_add(struct io_kiocb *req)
 {
-	struct io_uring_task *tctx = req->task->io_uring;
+	struct io_uring_task *tctx = req->tctx;
 	struct io_ring_ctx *ctx = req->ctx;
 
 	/* task_work already pending, we're done */
@@ -1226,7 +1223,7 @@ static void io_req_normal_work_add(struct io_kiocb *req)
 		return;
 	}
 
-	if (likely(!task_work_add(req->task, &tctx->task_work, ctx->notify_method)))
+	if (likely(!task_work_add(tctx->task, &tctx->task_work, ctx->notify_method)))
 		return;
 
 	io_fallback_tw(tctx, false);
@@ -1343,8 +1340,7 @@ static void io_req_task_cancel(struct io_kiocb *req, struct io_tw_state *ts)
 void io_req_task_submit(struct io_kiocb *req, struct io_tw_state *ts)
 {
 	io_tw_lock(req->ctx, ts);
-	/* req->task == current here, checking PF_EXITING is safe */
-	if (unlikely(req->task->flags & PF_EXITING))
+	if (unlikely(io_should_terminate_tw()))
 		io_req_defer_failed(req, -EFAULT);
 	else if (req->flags & REQ_F_FORCE_ASYNC)
 		io_queue_iowq(req);
@@ -1403,7 +1399,7 @@ static void io_free_batch_list(struct io_ring_ctx *ctx,
 		}
 		io_put_file(req);
 		io_req_put_rsrc_nodes(req);
-		io_put_task(req->task);
+		io_put_task(req);
 
 		node = req->comp_list.next;
 		io_req_add_to_cache(req, ctx);
@@ -2019,7 +2015,7 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
 	req->flags = (__force io_req_flags_t) sqe_flags;
 	req->cqe.user_data = READ_ONCE(sqe->user_data);
 	req->file = NULL;
-	req->task = current;
+	req->tctx = current->io_uring;
 	req->cancel_seq_set = false;
 
 	if (unlikely(opcode >= IORING_OP_LAST)) {

io_uring/io_uring.h

@@ -426,6 +426,19 @@ static inline bool io_allowed_run_tw(struct io_ring_ctx *ctx)
 		      ctx->submitter_task == current);
 }
 
+/*
+ * Terminate the request if either of these conditions are true:
+ *
+ * 1) It's being executed by the original task, but that task is marked
+ *    with PF_EXITING as it's exiting.
+ * 2) PF_KTHREAD is set, in which case the invoker of the task_work is
+ *    our fallback task_work.
+ */
+static inline bool io_should_terminate_tw(void)
+{
+	return current->flags & (PF_KTHREAD | PF_EXITING);
+}
+
 static inline void io_req_queue_tw_complete(struct io_kiocb *req, s32 res)
 {
 	io_req_set_res(req, res, 0);

io_uring/msg_ring.c

@@ -89,8 +89,8 @@ static void io_msg_tw_complete(struct io_kiocb *req, struct io_tw_state *ts)
 static int io_msg_remote_post(struct io_ring_ctx *ctx, struct io_kiocb *req,
 			      int res, u32 cflags, u64 user_data)
 {
-	req->task = READ_ONCE(ctx->submitter_task);
-	if (!req->task) {
+	req->tctx = READ_ONCE(ctx->submitter_task->io_uring);
+	if (!req->tctx) {
 		kmem_cache_free(req_cachep, req);
 		return -EOWNERDEAD;
 	}

io_uring/notif.c

@@ -89,7 +89,7 @@ static int io_link_skb(struct sk_buff *skb, struct ubuf_info *uarg)
 
 	/* make sure all noifications can be finished in the same task_work */
 	if (unlikely(notif->ctx != prev_notif->ctx ||
-		     notif->task != prev_notif->task))
+		     notif->tctx != prev_notif->tctx))
 		return -EEXIST;
 
 	nd->head = prev_nd->head;
@@ -115,7 +115,7 @@ struct io_kiocb *io_alloc_notif(struct io_ring_ctx *ctx)
 	notif->opcode = IORING_OP_NOP;
 	notif->flags = 0;
 	notif->file = NULL;
-	notif->task = current;
+	notif->tctx = current->io_uring;
 	io_get_task_refs(1);
 	notif->file_node = NULL;
 	notif->buf_node = NULL;

io_uring/poll.c

@@ -224,8 +224,7 @@ static int io_poll_check_events(struct io_kiocb *req, struct io_tw_state *ts)
 {
 	int v;
 
-	/* req->task == current here, checking PF_EXITING is safe */
-	if (unlikely(req->task->flags & PF_EXITING))
+	if (unlikely(io_should_terminate_tw()))
 		return -ECANCELED;
 
 	do {

io_uring/rw.c

@@ -435,7 +435,7 @@ static bool io_rw_should_reissue(struct io_kiocb *req)
 	 * Play it safe and assume not safe to re-import and reissue if we're
 	 * not in the original thread group (or in task context).
 	 */
-	if (!same_thread_group(req->task, current) || !in_task())
+	if (!same_thread_group(req->tctx->task, current) || !in_task())
 		return false;
 	return true;
 }